Everything in the game is invented: the town, the reports, the comments and the headlines. It is a teaching and research prototype with no efficacy data.
Everything in the game is invented: the town, the reports, the comments and the headlines. It is a teaching and research prototype with no efficacy data.
Each round asks a yes/no question about the invented town of Ardley. Two lines are evidence: a base rate ("about 30 in 100 calls like this turn out real") and a report with a stated strength ("2× as likely if it is real"). Bayes' rule turns those into one right answer. Everything else on the card, a comment thread, a drone, an earlier headline, a grievance, a countdown, carries no evidence.
From your answer the game infers λ̂, the weight you put on the report. At 1 you moved as far as the report justifies; above 1 further; between 0 and 1 less far; below 0 the wrong way. Each loaded round has a neutral twin with identical numbers, so the gap between the two is the part of your answer the feeling moved, not your arithmetic.
Part 2 adds three quick steps before each answer: place your feeling on a grid of pleasant to unpleasant and calm to activated, give it a name, and say what on the screen, or in you, produced it. Two of Part 2's pulls are new, to see whether the habit carries over.
This page used to hold a short game in the style of Bad News, the Cambridge and DROG game in which "players take on the role of a fake news producer and learn to master six documented techniques" (Roozenbeek and van der Linden 2019). Go Viral! and Harmony Square follow the same inoculation design. Borrowed Feelings departs from it on mechanism.
Why the scoring matters: a reanalysis of five studies found that Bad News and Go Viral! "did not improve discrimination, but rather elicited more 'false' responses to all news items (more conservative responding)" (Modirrousta-Galian and Higham 2023). Here every report is true and correctly stated, so blanket suspicion scores as badly as over-reaction.
It is a five-minute demonstration of an idea from Jonathan Walberg's working papers: that emotion changes how much weight people give evidence, and that a campaign can exploit this without faking anything. The λ form comes from the Sonic Vectors design and the Emotional Priors papers. The Emotional Priors paper predicts that teaching people "to notice when their reactions are being moved" should transfer across topics, while topic-specific hardening should not (its hypothesis H4). It also flags whether people who correctly attribute their arousal are still moved as the model's "most exposed seam." Part 2 is a toy version of both questions.
It isn't a validated measure or an intervention with evidence behind it. There are no efficacy data. Three twin pairs per part is far too few to say anything about a person; Part 2 always follows Part 1, so practice alone could shrink the gap; and the score mixes numeracy with feeling. In Walberg's simulation the transfer of affect-awareness follows from how awareness is specified, so it is a prediction, not a result.
Nothing you enter leaves your browser. The link stores only your answers.
Walberg's working papers (unpublished)
Prior games this one departs from
Affect, attribution and labeling
Design note: tools/prebunking-game/DESIGN.md. Round content: data/rounds.js.